RLMBLOCK

Recursive Language Model patterns for Claude Code — handle massive contexts (10M+ tokens) by treating them as external variables.

Based on: https://arxiv.org/html/2512.24601v1

How Users Interact

You don't call RLM tools directly. You ask Claude to analyze large files, and Claude uses RLM behind the scenes.

Example:

• You say: "Analyze this 2MB log file for errors"
• Claude uses RLM tools internally
• You get: "I found 3 error patterns: database timeouts (47), auth failures (23)..."

Core Idea

Instead of feeding massive contexts directly into the LLM:

1. Load context as external variable (stays out of prompt)
2. Inspect structure programmatically
3. Chunk strategically (lines, chars, or paragraphs)
4. Sub-query recursively on chunks
5. Aggregate results for final synthesis

Quick Start

Installation

git clone https://github.com/richardwhiteii/rlm.git
cd rlm
uv sync

Or with pip:

python -m venv .venv
source .venv/bin/activate
pip install -e .

Configure for Claude Code

Option 1: Quick Setup (recommended)

# From the rlm directory
claude mcp add rlm -s user -- uv run --directory "$(pwd)" python -m src.rlm_mcp_server

This adds RLM globally (-s user) so it's available in all your Claude Code sessions.

Option 2: With Ollama (free local inference)

First set environment variables, then add:

export RLM_DATA_DIR="$HOME/.rlm-data"
export OLLAMA_URL="http://localhost:11434"

claude mcp add rlm -s user -- uv run --directory "$(pwd)" python -m src.rlm_mcp_server

Option 3: Manual JSON config

Add to ~/.claude/.mcp.json for full control:

{
"mcpServers": {
"rlm": {
"command": "uv",
"args": ["run", "--directory", "/path/to/rlm", "python", "-m", "src.rlm_mcp_server"],
"env": {
"RLM_DATA_DIR": "/path/to/.rlm-data",
"OLLAMA_URL": "http://localhost:11434"
}
}
}
}

Note: Replace `/path/t